My team is attempting to calculate conversation metrics in Power BI, using the data piped into our data lake. We’re attempting to count the number of replies across all conversations over a particular time period. I know the part_author is “Admin,” but what conversation_part values constitute a reply? We want to match the metrics provided via the in-app reporting tool.
Hey there
To calculate the number of replies across all conversations over a particular time period, you should focus on conversation parts where the part_type is comment. This indicates a reply made by a teammate or admin.
Here’s how you might proceed:
- Filter Conversation Parts: Look for conversation parts with part_type = comment and ensure they fall within your specified date range
- Count Unique Conversations: Count the unique number of conversations that include these parts to get the number of replies
Please note that the count_conversation_parts under statistics includes all conversation parts, including the initial message, while conversation_parts.total_count excludes the first message. This distinction is important when calculating replies, as you should focus on parts beyond the initial message.
Hope this helps! 💪🏼
Thank you!
No problem
Hi
Thanks for the helpful explanation above — it clarified a lot for us too!
We’ve been digging into our Intercom data via the API and noticed something interesting: while comment
is clearly the primary part_type
for replies, we’ve also come across some other part_type
s that include teammate replies — like assign_and_reopen
, open
, or assignment
— even though they’re not technically classified as comment
.
From what we can tell, this happens when a teammate adds a message while (re)assigning and/or (re)opening the conversation.
So while we understand that the official guidance for calculating replies focuses on part_type = "comment"
, it seems like some actual replies might be missed if we don’t also consider other part_type
s that include a non-empty body
.
Have you seen this happen before? Would love to know if this is an edge case worth flagging — or something specific to how our workspace is using Intercom.
Thanks again!
Join the Intercom Community 🎉
Already have an account? Login
Login to the community
No account yet? Create an account
Intercom Customers and Employees
Log in with SSOEnter your E-mail address. We'll send you an e-mail with instructions to reset your password.